applying gcms to terrestrial exoplanets f.forget, r. wordsworth b. charnay, e. millour, f. codron,...

Applying GCMs to terrestrial exoplanets

F.Forget, R. WordsworthB. Charnay, E. Millour, F. Codron, J.Leconte,

F. Selsis*,

LMD, Institut Pierre Simon Laplace, Université Pierre et Marie Curie, BP 99, 75005 Paris, France* LAB, (CNRS; Université Bordeaux I)

MercuryVenus

Earth

Mars

JUPITER

SATURNETitanURANUS

NEPTUNETritonAtmospheres in the solar system

• GIANT PLANETS• Terrestrial atmospheres

Pluto

VENUS

TERRE

MARS TITAN

Many GCM teamsApplications:• Weather forecast• Assimilation and climatology• Climate projectionsClimate projections• Paleoclimates• chemistry• Biosphere / hydrosphere cryosphere / oceans coupling• Many other applications

~a few GCMs (LMD, Univ. of Chicago, Caltech, Köln…)

Coupled cycles:• Aerosols• Photochemistry• Clouds

Several GCMs (NASA Ames, Caltech, GFDL, LMD, AOPP, MPS, Japan, York U., Japan, etc…)Applications:• Dynamics & assimilation• CO2 cycle• dust cycle• water cycle• Photochemistry• thermosphere and ionosphere• isotopes cycles• paleoclimates• etc…•

~1 true GCMs Coupling dynamic & radiative transfer(LMD)6 GCMs with simplified physics

TRITON1 full GCM (LMD)

PLUTO1 full GCM (LMD)

What we have learned from solar system GCMs

• To first order: GCMs work– A few equations can build « planet simulators » with a realistic, complex

behaviour and strong prediction capacities

• However the devil is in the details:– Problems with

• Positive feedbacks and unstability (e.g. sea ice and land ice albedo feedback on the Earth)

• Non linear behaviour and threshold effect (e.g. dust storms on Mars)

• Complex subgrid scale process and poorly known physics (e.g. clouds on the Earth)

• System with extremely long « inertia » with small forcing (e.g. Venus circulation). Sensitivity to intrinsinc dynamical core conserving properties Sensitivity to initial state

Need to somewhat « tune » a few model parameters to accurately model an observed planet and predict its behaviour

Lebonnois et al. 2011

Venus intercomparison GCMs

Lebonnois et al. 2011

Lebonnois et al. 2011

Lebonnois et al. 2011

Lebonnois et al. 2011

Mars dynamical core intercomparisons

(Mischna and Wilson 2008)

Participating Model SpecsModel Core numerics Vertical Levels

GFDL A Spectral 28, 36

GFDL B B-grid 32

GFDL C Finite Volume 28, 36

GFDL D FV + Cubed-Sphere 28, 36, 56

Hokkaido Spectral 48

Caltech/Cornell/JPL C-grid 25, 40

CCSR/NIES Spectral 33

York C-grid 88

LMD C-grid 32

Various models provided various combinations of model resolution, damping (α=2, 4) and vertical levels.

Hokkaido Spectral MGCM

GFDL Spectral MGCM: T31L36

GFDL B-grid MGCM

GFDL Cubed-sphere MGCM 3°x3°

Caltech/Cornell/JPL MarsWRF MGCM

LMD MGCM

York University MGCM

Full intercomparison

Johnson et al. 2008

The science of Simulating the unknown:From planet GCMs to extrasolar planet GCMs.

How GCM work ? :The minimum General Circulation Model for a terrestrial planet

1) 3D Hydrodynamical code to compute large scale atmospheric motions and transport

2) At every grid point : Physical parameterizations to force the dynamic to compute the details of the local climate• Radiative heating & cooling of the atmosphere • Surface thermal balance • Subgrid scale atmospheric motions

Turbulence in the boundary layer Convection Relief drag Gravity wave drag• Specific process : ice condensation, cloud microphysics, etc…

Most processes can be described by equations that we have learned to solve with some accuracy

Toward a “generic” Global climate model (LMD)

1) Use “universal” parametrisations for all planets:

– Standard dynamical core– Surface and subsurface thermal model– “Universal” Turbulent boundary layer scheme

2) The key : Versatile, fast and accurate radiative transfer code (see next slide)

3) Simple, Robust, physically based parametrisation of volatile phase change processes

- Including robust deep convection representation : wet convection- Clouds are represented by splitting condensed phase on a prescribed number of

Cloud Condensation Nuclei (see next slide)

4) If needed: simplified physical “slab ocean + sea ice” scheme.

Developping a Versatile, fast and accurate radiative transfer code for GCM

• Input : assumption on the atmosphere:– Any mixture of well mixed gases (ex: CO2 + N2 + CH4 + SO2)– Add 1 variable gases (H2O). Possibly 2 or 3 (e.g. Titan)– Refractive Indexes of aerosols.

• Semi automatic processes: Spectroscopic database (Hitran 2008) Line by line spectra (k-spectrum model) correlated k coeeficients radiative transfer model

• RT Model can also simulate scattering by several kind of aerosols (size and amount can vary in space and time)

• Key technical problems: – gas spectroscopy in extreme cases– Predicting aerosol and cloud properties

• Key scientific problem : assumption on the atmosphere !

Clouds and precipitation (at this stage)

• Large scale condensation:

(r~.0.2 …)

• Wet convection (“Manabe”) mass flux scheme ?

• Particles size estimated by splitting condensed phase on a prescribed number of Cloud Condensation Nuclei. Used to compute:– Sedimentation– cloud effective radius

• Precipitation : above threshold:

ql/(cloud fraction) > threshold (~1.1g/kg)

• On the surface : « bucket model » (runoff if > 150 kg/m2) + snow, etc.

1-rqH2O qH2O 1+rqH2O

Cloud fraction

An ongoing test: modeling the Earth as an extrasolar planet

• An « Earth size » planet at 1 AU around a G2 star…

• Asume N2 atmosphere + present-day CO2 and CH4

• Use Earth albedo and topography.• Oceans are represented by humid, very

high thermal inertia surface (TI=25000 SI).• Sea ice limited to 1 meter thick

To be replaced by a sea –ice model.

26

27

Reference(Earth climatology LMDZ4)

Simulation with the Exoplanet generic model

Mean surface temperature (K)

Mean surface temperature (K)

64x64x20 simulations

Reference(Earth climatology LMDZ4)

Simulation with the Exoplanet generic model

— Earth reference— GCM Generic

29

Reference(Earth climatology LMDZ4)

Simulation with the Exoplanet generic model

Mean precipitation (kg m2 s-1)

Mean precipitation (kg m2 s-1)

30 We need a better ocean transport and sea ice model !

A new « slab ocean » model developped at LMD

Codron, Climate Dynamics, 2011

• How to simply represent Oceanic transport?– Empirical flux based on terrestrial observations ?

– Horizontal diffusion of temperature ?

– Parametrize the largest heat transport : wind-driven Ekman transport (and return flow at depth)

31

32

Test :• Aquaplanet• Earth with LMDZ GCM

Codron, Climate Dynamics, 2011

An example of application : Gliese 581d

Star = 0.31 MsunDistance = 20.4 Light Year

Planet C :M > 5 Earth MassOrbit : 13 daysat 0.073 AU

Planet B : 16 M

Planet D :M > 7. Earth MassOrbit : at 0.25 AU Bonfils et al. 2005

Udry et al. 2007Selsis et al. 2007

Gliese 581 system

GGliese 581d

• Mass ~ < 7 Earth Mass• Orbit around a small cold M star

– 1 year ~ 67 Earth days– Excentricity ~ 0.38– Equilibrium temperature ~ -80°C

• Gravity : between 10 and 30 m s-2

• Tidal forces :– Possibly locked in synchronous rotation

or a resonnance– Most likely low obliquity (like Venus or Mercury)

A GCM for Gliese 581d• The question: what could be the climate assuming a CO2 –

N2 – H2O atmosphere ? Could Gliese 581d be habitable ?• The model :

– Two spatial resolution • 11.25° lon x 5.6° lat resolution• 5.6° lon x 2.3° lat resolution

– Radiative transfer : • 32 spectral bands in the longwave and 36 in the shortwave • Include improved Collision Induced absorption parametrisation (Wordsworth et

al. 2010)

• 6 × 9× 7 temperature, pressure and H2O volume mixing ratio grid: – T =100; 150;…350K, – p = 1E-3, 1E-2, … 1E5 mbar

– qH2O = 1E-7, 1E-6 … 1E-1

• 16 points in the g space (k-distributions)

• Assume stellar spectra from Virtual planet laboratory (AD Leo, slightly warmer than Gliese 581)

– CO2 condensation (surface + clouds) included

– Two cases : 1) desert planet 2) Ocean planet (not oceanic transport)

top of atmGround, no Rayleigh scat

Ground, with Rayleigh scat.

top of atmGround, no Rayleigh scat

Ground, with Rayleigh scat.

G-class spectrum(Sun)

M-Class spectrum(AD Leo and ~Gliese 581)

Incident solar flux on a 40 bars CO2 atmosphere

Rayleigh scattering

CO2 absorption

Wordsworth et al. 2010

Global Mean results (1D radiative convective models)

Wordsworth et al. (A&A , 2010)

M star (AD Leo)

SunMe

an

Su

rfa

ce

Te

mp

era

ture

(K

)

Surface Pressure (bars)

From 1D to 3D

• Distribution of clouds

• Water cycle : cold trapping of water ?

• Impact of heterogeneous heating (cold hemisphere, poles)

Tidal locked Gl581d Ps=10barSurface Temperature

Resonnant 2/1 Gl581d Ps=10barSurface Temperature

Earth-like rotation Gl581d Ps=10barSurface Temperature

Mean surface temperature and atmospheric CO2 collapse

10 bars

CO2 Liquid !

deposition

Simple CO2 ice cloud scheme

1. In each model mesh: If T<Tcond : condensation and latent heat release T=Tcond

2. CO2 ice is splitted in small particles (The number of particle / kg is prescribed)

3. Transport and mixing by winds, turbulence, convection

4. Gravitional sedimentation5. Interaction with Solar and IR radiation (assuming

Mie theory and Hansen et al. (1996) radiative properties

6. If T>Tcond : sublimation to get T=Tcond or no more ice

CO2 ice clouds maps: Res 2/1 gl581bIce condense in ascendance (adiabatic cooling)

CO2 ice clouds maps: tidal locked gl581b

Tidal locked Gl581d Ps=20barSurface Temperature

Mean surface temperature (No atmospheric CO2 collapse)

20 bars

Cloud condensation

Radiative effects (clouds + vapour)

Precipitation

Surface processes

Evaporation

• We include radiative effects of vapour and cloud tracers

• Assume fixed CCN distribution, but variable mean cloud particle sizes

• Simple convective relaxation (Manabe scheme), 100% cloud fraction assumed

Adding a water cycle

Ocean planet (no oceanic transport)Mean cloud coverage

H2O CO2

ice ice

Including the water cycle:Preliminary results

(assuming an ocean planet)

Including the water cycle:Preliminary results

Dry case

Water cycle case

• Net warming• No glaciation with Ps > 10 bars

Assuming enough CO2 and H2O (which is not unlikely), Gliese 581d WOULD be habitable.

Gliese 581d: conclusions

Dry planet

With water cycle

Astrophysical Journal, Wordsworth, Forget et al. 2011

Terrestrial planets: Many projects• Cold atmosphere

– Early Mars– Early Earth– Cold ocean planet with oceanic transport and sea ice

• Warm atmosphere– Habitable zone Inner edge : runaway greenhouse effect, etc…

Challenge: clouds+convection+water vapor radiative properties– Early Venus– Future Earth

• Exotic cases:– H2 atmospheres (Wordsworth 2011)

Some Conclusions• Assuming atmosphere/ocean compositions, robust,

“complete” GCMs/Planet simulators may be developed to address major scientic questions related to extrasolar planets :– Prepare observations (e.g. F. Selsis yesterday)– Could such and such planet be habitable. – Limits of habitability

– However, whatever the quality of the model, heavy study of model sensitivity to parameters will always be necessary.

• The key scientific problem may be to understand the zoology of atmospheric composition and long term evolution… observations by a project like ECHO would tell us a lot.

• To be continued…